Blanc et al. Genet Sel Evol (2021) 53:24 https://doi.org/10.1186/s12711-021-00614-5 Genetics Selection Evolution

RESEARCH ARTICLE Open Access Infuence of genetics and the pre‑vaccination blood transcriptome on the variability of antibody levels after vaccination against Mycoplasma hyopneumoniae in pigs Fany Blanc1* , Tatiana Maroilley1 , Manuel Revilla1 , Gaëtan Lemonnier1, Jean‑Jacques Leplat1, Yvon Billon2, Laure Ravon2, Olivier Bouchez3, Jean‑Pierre Bidanel1 , Bertrand Bed’Hom1 , Marie‑Hélène Pinard‑van der Laan1 , Jordi Estellé1 and Claire Rogel‑Gaillard1*

Abstract Background: The impact of individual genetic and genomic variations on immune responses is an emerging lever investigated in vaccination strategies. In our study, we used genetic and pre-vaccination blood transcriptomic data to study vaccine efectiveness in pigs. Results: A cohort of 182 Large White pigs was vaccinated against Mycoplasma hyopneumoniae (M. hyo) at weaning (28 days of age), with a booster 21 days later. Vaccine response was assessed by measuring seric M. hyo antibodies (Ab) at 0 (vaccination day), 21 (booster day), 28, 35, and 118 days post-vaccination (dpv). Inter-individual variability of M. hyo Ab levels was observed at all time points and the corresponding heritabilities ranged from 0.46 to 0.57. Ab per‑ sistence was higher in females than in males. Genome-wide association studies with a 658 K SNP panel revealed two genomic regions associated with variations of M. hyo Ab levels at 21 dpv at positions where immunity-related genes have been mapped, DAB2IP on chromosome 1, and ASAP1, CYRIB and GSDMC on chromosome 4. We studied covari‑ ations of Ab responses with the pre-vaccination blood transcriptome obtained by RNA-Seq for a subset of 82 pigs. Weighted gene correlation network and diferential expression analyses between pigs that difered in Ab responses highlighted biological functions that were enriched in heme biosynthesis and platelet activation for low response at 21 dpv, innate antiviral immunity and dendritic cells for high response at 28 and 35 dpv, and cell adhesion and extra‑ cellular matrix for high response at 118 dpv. Sparse partial least squares discriminant analysis identifed 101 genes that efciently predicted divergent responders at all time points. We found weak negative correlations of M. hyo Ab levels with body weight traits, which revealed a trade-of that needs to be further explored. Conclusions: We confrmed the infuence of the host genetics on vaccine efectiveness to M. hyo and provided evidence that the pre-vaccination blood transcriptome co-varies with the Ab response. Our results highlight that both

*Correspondence: [email protected]; [email protected] 1 Université Paris‐Saclay, INRAE, AgroParisTech, GABI, 78350 Jouy‐en‐Josas, France Full list of author information is available at the end of the article

© The Author(s) 2021. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativeco​ ​ mmons.org/licen​ ses/by/4.0/​ . The Creative Commons Public Domain Dedication waiver (http://creativeco​ mmons​ .org/publi​ cdoma​ in/​ zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Blanc et al. Genet Sel Evol (2021) 53:24 Page 2 of 18

genetic markers and blood biomarkers could be used as potential predictors of vaccine response levels and more studies are required to assess whether they can be exploited in breeding programs.

Background for respiratory syncytial virus vaccine in bovine [24] and Sustainability is one of the main current challenges in for Newcastle disease virus vaccine in chicken [25]. In livestock farming. In this context, reducing the use of pigs, across-breed variability of response levels to vac- antibiotics and anti-microbials has become a major con- cine against Aujeszky’s virus disease has been reported cern. Tis can be addressed by considering that animals [26]. Within-population variability has been shown for continually interact with a dynamic and potentially path- Ab response to vaccines against infuenza [27] and por- ogenic ecosystem. Increasing vaccination efciency is cine reproductive and respiratory syndrome [28] viruses, one of the avenues explored to promote sustainable live- and tetanus bacteria [29], and to the bacterial antigens stock production. To achieve this, it is necessary to better K88ab, K88ac, and O149 [30]. We and others have also understand the mechanisms that underlie host-patho- reported individual variability of serum M. hyo-specifc gens interactions, to develop new vaccines and vaccina- Ab induced after vaccination in various populations [31, tion strategies, and also to consider genetic improvement 32]. of the host response to vaccination and management of Blood is an accessible fuid that refects the status of the its variability [1, 2]. immune system [33]. Tus, in vaccine research, it serves Mycoplasma hyopneumoniae (M. hyo) bacteria are as a surrogate tissue to identify potential markers of vac- known to cause enzootic pneumonia, a chronic respira- cine-induced responses. In swine, we have shown that tory disease in pigs, and play a primary role in the por- the peripheral blood transcriptome refects variations cine respiratory disease complex [3]. M. hyo infections in innate and adaptive immunity traits [34]. In humans, cause signifcant economic losses due to the costs of blood transcriptomic profling has identifed both early treatments, reduced animal performance by decreasing innate [35, 36] and concomitant adaptive humoral immu- growth, and increased mortality from secondary infec- nity gene signatures after vaccination [37]. In particular, tions [4]. Commercial vaccines are efective in prevent- Li et al. [38] defned early transcriptional signatures of Ab ing and reducing the severity of lung lesions, and thus in responses that were derived from a systems biology study. improving daily weight gain and slaughter weight [5–8]. Such studies have also been recently conducted in sheep However, they do not reduce the transmission of the [39, 40] and pigs [41–43], and have provided insights into pathogen signifcantly [7, 9]. Terefore, improving vacci- the pathways involved in vaccination responses. How- nation efcacy to M. hyo is still an important issue. ever, to date, the underlying determinants of immune Systems vaccinology and vaccinomics, which con- capacity that are involved in vaccination responses have sist of merging -omics data to comprehensively assess not been identifed. biological systems in response to vaccination, are new In this study, our aim was to characterize individual approaches that have been proposed in human vaccinol- variability of vaccine response to M. hyo in pigs and to ogy to enhance insights into the evaluation of immune identify genetic parameters and baseline blood transcrip- responses and adverse events, and the development of tomic profles that could predispose to efective response new vaccine candidates [10, 11]. Immunogenetic stud- to vaccination and predict associated Ab response levels. ies in humans have revealed that single nucleotide poly- Tus, we combined data from high-density genotyping, morphisms (SNPs) in human leukocyte antigen class the transcriptome of blood collected before vaccination, I and class II, cytokine, cytokine receptor, and innate and M. hyo-specifc Ab levels at various time points after immune response (e.g., toll-like receptor) genes may vaccination. partly account for the inter-individual variability of the immune response to various vaccinations (measles and Methods rubella [1, 12–16], hepatitis B [1, 17–19], infuenza [20], Animal design, zootechnical traits, and sampling smallpox [21] or Bacillus anthracis [22]). In humans, In total, 48 litters of Large White pigs were produced in host genetics has been reported to infuence the persis- fve batches and raised without antibiotic treatment on tence of specifc antibodies (Ab) throughout life, after the GenESI, INRAE, Pig Innovative Breeding Experi- vaccination against tetanus toxoid, infuenza B virus, and mental Facility (https​://doi.org/10.15454​/1.55724​15481​ capsular group C meningococcal during childhood [23]. 18584​7E12). Sows (n = 47) were inseminated with boar In livestock animals, the role of host genetics in the vari- semen (n = 48) that was selected to maximize genetic ability of Ab response to vaccines has been documented variability, with one sow inseminated for two parities. Blanc et al. Genet Sel Evol (2021) 53:24 Page 3 of 18

From each litter, three to four piglets of each sex were death (n = 13 vaccinated and 11 controls) or to no chosen based on their weight at 21 days, taking care to response to vaccination during the experiment (n = 4). represent average litter weight piglets and avoid animals Te latter four animals were not included because we with a too low weight. Tis resulted in a set of 278 piglets could not conclude whether they were non-responders or (145 uncastrated males and 133 females) that defned the whether the vaccination injection had failed. experimental population. At random four to fve piglets Te animals were raised under standard conditions in per litter were vaccinated against M. hyo (Stellamune, pens of 20 to 30 animals during the post-weaning period Pfeizer), and the remaining piglets (one or two animals (from around 28 to 68 days of age) and in pens of 10 to per litter) were not vaccinated. In total, the experimen- 12 animals during the growth period (from around 68 tal population included 203 vaccinated animals and 75 to 146 days of age). Tey all received the same stand- control non-vaccinated animals. Te experimental design ard commercial diets. Te fnal dataset comprised 186 with the associated measures at diferent time points is piglets vaccinated against M. hyo and 64 control non- summarized in Fig. 1a. A frst injection was administered vaccinated piglets, of which 170 and 61, respectively, at 0 dpv, corresponding to the day of weaning (at 28 days were monitored until 118 dpv. Te fnal individual data- of age on average, from 24 to 31 days). A booster vacci- set is in Additional fle 1: Table S1. Body weights (BW) nation injection was given at 21 dpv. A small subset of were determined at diferent time points: at birth, 7 days animals was removed from the initial population due to before weaning (dbw) at around 21 days of age, at wean- farming problems that led to morbidity and premature ing at around 28 days of age (corresponding to 0 dpv), at

a

b cd

Fig. 1 Individual variability of M. hyo Ab response in pigs’ sera after vaccination. a Vaccination, blood sampling protocol and associated data. The time points for body weight measures and blood samples are represented by stars. b Levels of M. hyo Ab in 1/40 diluted sera from vaccinated and non-vaccinated animals at 0 (day of vaccination), 21 (day of booster), 28, 35, and 118 days post vaccination (dpv). Dotted lines represent thresholds of the assay given by the provider. Statistical analyses: unpaired t test between controls and vaccinated animals (***p < 0.001). c Levels of M. hyo Ab in 1/4 diluted sera from vaccinated and control animals at 21 dpv. Dotted lines represent thresholds of detection (mean 2 SD of + the controls). Statistical analyses: unpaired t test between controls and vaccinated animals (***p < 0.001). d Levels of M. hyo Ab in 1/40 diluted sera from vaccinated males and females at 0, 21, 28, 35, and 118 dpv. Dotted lines represent the thresholds of the assay given by the provider. Statistical analyses: data were ftted using a mixed linear regression model with sex, age at weaning and batch as fxed efects, and litter as random efect; and efect of sex was evaluated by ANOVA (**p < 0.01) Blanc et al. Genet Sel Evol (2021) 53:24 Page 4 of 18

the end of the post-weaning period (corresponding to and 0.545 at 1/4 dilution). Animals for which values were 40 dpv) at around 68 days of age, and at the end of the below vs above these thresholds at both dilutions were experiment, before slaughter (corresponding to 118 dpv) classifed as negative (n = 134) vs positive (n = 32). Ani- at around 146 days of age. Average daily gain (ADG) was mal groups at each time point are in Additional fle 1: calculated from 0 to 40 dpv (ADG 0–40 dpv) and from 40 Table S1 and descriptive statistics of the groups are in to 118 dpv (ADG 40–118 dpv). Additional fle 1: Table S2. Peripheral blood was sampled from the jugular vein at diferent time points using dry tubes (Becton Dickinson) Efect of zootechnical parameters on vaccination response for serum preparation (0, 21, 28, 35, and 118 dpv), EDTA- and of vaccination on body weight coated tubes (Becton Dickinson) for DNA extraction (0 Unless otherwise notifed, all data analyses were con- dpv), and Tempus tubes (Termo Fisher) for RNA extrac- ducted with the R software (v3.6.1) [44]. Te efects of dif- tion (0 dpv). Blood samples were stored at − 20 °C (EDTA ferent zootechnical parameters on M. hyo Ab responses tubes) or at − 80 °C (Tempus tubes) prior to DNA or were evaluated with a linear mixed model using the lmer RNA extraction, respectively. Peripheral blood was also function of the lme4 (v1.1-21) R package [45], with sex sampled from sows on dry tubes during the week before (two levels) and batch (fve levels) as fxed efects, age parturition to check that they were seronegative for M. at weaning (between 24 and 31 days) as a linear covari- hyo-specifc Ab. ate, and litter (48 levels) as a random efect. To evaluate the efect of vaccination on weight traits, vaccine groups Measurement of M. hyo‑specifc Ab levels and classifcation (vaccinated vs non-vaccinated; or high vs low; or nega- of animals based on Ab response tive vs positive) were also included in the model as fxed Te levels of M. hyo-specifc Ab were measured by using efects. P-values for fxed efects were obtained using the a commercial ELISA test (IDEXX M. hyo Ab test, IDEXX lmerTest package (v3.1-1) [46] by type III ANOVA tables, Europe B.V., Te Netherlands) and running duplicates with Satterthwaite’s approximation to degrees of freedom of sera diluted 1/40 for all time points (0, 21, 28, 35, and using the ANOVA function, while p-values for random 118 dpv). At 21 dpv, M. hyo-specifc Ab were also meas- efects were calculated by the likelihood ratio test using ured with sera diluted 1/4. Ab levels were calculated by the rand function. Pairwise comparisons with Tukey’s dividing the absorbance of the samples (S, corrected by adjustment were performed to assess the diferences subtraction of the mean negative control absorbance) by between batches, using the emmeans function of the the mean absorbance of the positive control (P, corrected emmeans package (v1.5.2-1). A signifcance level of 0.05 by subtraction of the mean negative control absorbance), was applied. Pearson’s correlation matrices were built which resulted in an S/P value following the IDEXX pro- with the corrplot R package (v0.84). Correlations of the cedure. Given the thresholds provided by IDEXX, sam- repeated measures of Ab response levels at the diferent ples with S/P values higher than 0.4, within the range time points post-vaccination were also calculated using from 0.3 to 0.4, or lower than 0.3, were assessed as posi- a linear mixed model with dpv as fxed efect, animal as tive, suspect, or negative for M. hyo Ab, respectively. random efect, and an autoregressive correlation struc- To carry out diferential analyses between animals with ture of order 1 (corCAR1 function) across dpv, using the high and low vaccine responses, we identifed extreme nlme package (v3.1-141). Principal component analyses groups for M. hyo-specifc Ab levels at 21, 28, 35, and (PCA) were performed with the FactoMineR R package 118 dpv, using the S/P values obtained for the whole pig (v2.1) [47] and the results were visualized based on the population at each time point (n = 186 at 21, 28, and 35 factoextra R package (v1.0.6). dpv and n = 170 at 118 dpv). High (low) responders cor- responded to pigs with an Ab response higher (lower) Estimation of heritabilities than the mean plus (minus) one standard deviation Genetic parameters (heritabilities and genetic correla- (SD) (n = 34, 30, 22, and 27 for high responders, colored tions) were estimated using the restricted maximum in blue in Fig. 1a, b, and n = 32, 29, 28, and 25 for low likelihood methodology applied to a multivariate mixed responders, colored in orange in Fig. 1a, b at 21, 28, linear model with the same fxed efects and covariates 35, and 118 dpv, respectively). Positive (pos) and nega- (sex, batch and age at weaning) as above, and the additive tive (neg) animals were defned at 118 dpv based on the genetic value of each animal and a residual error as ran- thresholds of the Ab test given by the manufacturer (neg: dom efects, which is the so-called animal model, where S/P value < 0.3, n = 19, pos: S/P value > 0.4, n = 136). For the vector of additive genetic values is assumed to be Ab response at 21 dpv, the mean plus 2 SD of the values proportional to the numerator relationship matrix built obtained for non-vaccinated control animals were used from pedigrees. Computations were performed using the to determine detection thresholds (0.07 at 1/40 dilution VCE6 software [48]. Blanc et al. Genet Sel Evol (2021) 53:24 Page 5 of 18

Genome‑wide association studies (GWAS) at 94 °C. For synthesis of the frst strand, 1 µL of Super- Vaccinated pigs (n = 186) were genotyped with a high- Script II Reverse Transcriptase was mixed with 9 µL of density SNP panel (Afymetrix AXIOM PIG HD, 658 K Illumina’s First Strand Mix with actinomycin D, and then SNP). Genotyping of four pigs (2.1%) did not pass various 8 µL of this mix were added to the fragmented RNA, and internal quality controls (QC) that identify poor qual- PCR was carried out in a thermocycler that was pro- ity samples using the Axiom Suite [Dish QC (n = 2) and grammed as indicated in the TruseqRNA protocol. To sample QC call rate test (n = 2)]. Among the genotypes generate double strand (ds) cDNA, 20 µL of Illumina’s for the 182 remaining animals that passed plate QC (see Second Strand mix were mixed with the frst strand Additional fle 1: Table S1), only annotated autosomal cDNA and incubated for one hour at 16 °C. Ten, 90 µL SNPs were kept for analysis (n = 598,138; Axiom_PigHD_ of AMPure XP beads were used to purify the ds cDNA v1.na35.r4.a2.annot.csv annotation fle). Te check. that was eluted in 15 µL of Illumina’s resuspension bufer. marker function of the GenABEL package (v1.8-0) in Te ds cDNA was end-repaired, adenylated, and then R was applied to flter out SNPs with a minor allele fre- Illumina adapters were added, as indicated in the TruSeq quency lower than 5% (89,175 SNPs excluded), a call rate stranded mRNA protocol. Te prepared libraries were lower than 95% (58,869 SNPs excluded), or SNPs that quality-checked with the high sensitivity D1000 screen departed from Hardy–Weinberg equilibrium (FDR lower Tape (Agilent Tape Station 2200), quantifed with Qubit than 0.1; 41,149 SNPs excluded). After applying these QC (TermoFisher), and 12-plex pooled. Te pooled librar- measures, 425,567 SNPs on 182 animals were retained ies were quantifed with the Qubit dsDNA HS (High Sen- for GWAS. Overall, the QC analysis of the genotyping sitivity) Assay kit and sequenced on the GeT-PlaGe core data did not identify outlier animals and the genomic facility (INRAE, https​://doi.org/10.15454​/1.55723​70921​ kinship coefcients between individuals were consistent 30319​3E12) on the Illumina Hiseq3000 sequencer with with the known pedigrees. GWAS were performed with a 150PE module, with each pool run in two Hiseq3000 the RepeatABEL R package (v1.1) [49], by using a linear lanes. mixed model with batch and sex as fxed efects, age at Te reads were mapped to the pig genome assem- weaning as a linear covariate, and litter and the genomic bly Sscrofa11.1 (Ensembl v90 release) by using TopHat kinship matrix (built with the ibs function) as random (v2.1.0), and the read counts for each gene were obtained efects. A signifcance level of 0.05 and a suggestive sig- by using htseq_count (v0.6.1.p1). Overall, the RNA-Seq nifcance level of 0.1 were applied. Finally, the detected data provided a sufcient number of reads per sample 6 associated regions were mapped to the pig genome (mean number of reads larger than 60 × ­10 ). Tree ani- assembly available at the UCSC Genome Browser on Feb. mals with a total number of reads smaller than 20 million 2017 (Sscrofa11.1). were excluded. Tus, the fnal dataset for the transcrip- tome analysis included 82 vaccinated pigs, of which 15, Blood transcriptome by RNA‑Seq 14, 11, and 11 were high responders at 21, 28, 35, and 118 RNA from blood samples that were collected in Tem- dpv, respectively, 15, 23, 19, and 13 were low respond- pus tubes prior vaccination at 0 dpv was extracted from ers at 21, 28, 35, and 118 dpv, respectively, and 55 and 10 a subset of 85 vaccinated pigs (see Additional fle 1: were negative, and 14 and 64 were positive at 21 and 118 Table S1), using the Norgen Preserved Blood RNA Puri- dpv, respectively. Tese pigs were representative of the fcation Kit I (adapted to blood samples collected in corresponding groups from the whole population, show- Tempus tubes) according to the manufacturer’s instruc- ing equivalent means of Ab levels (see Additional fle 1: tions. Concentration of the extracted RNA was meas- Table S2). ured with a NanoDrop 2000 spectrophotometer (89.4 ± 26.0 µg were obtained per sample) and RNA integrity was Weighted gene correlation network analysis (WGCNA) assessed by an Agilent 2100 Bioanalyzer, using the eukar- and correlation with vaccine‑induced Ab responses yote total RNA 6000 Nano Kit (RIN obtained were 8.0 ± WGCNA [50] was conducted in R to fnd clusters of 0.6, ranging from 7.1 to 9.3). highly correlated genes within the RNA-Seq dataset. Libraries were prepared from 1 µg of total RNA with Read counts per gene were fltered by retaining only the Illumina TruSeq stranded total RNA with Ribo-Zero the genes with more than 1 count per million and more Globin sample preparation kit. Following the manu- than 10 reads for at least one third of the animals. Read facturer’s protocol, ribosomal and globin RNA were counts were then normalized with the calcNormFactors removed by depletion and the remaining coding and function implemented in the edgeR package (v3.26.6). non-coding RNA was used as input for library prepara- Finally, limma’s voom function (v3.40.6) was used to ft tion. RNA was fragmented using Illumina’s fragmenta- a generalized linear regression model to correct the data tion enzyme mix (Elute, Prime, Fragment Mix) for 8 min with sex and batch as fxed efects, and age at weaning Blanc et al. Genet Sel Evol (2021) 53:24 Page 6 of 18

as a linear covariate. Samples were then clustered based Sparse partial least square‑discriminant analysis (sPLS‑DA) on their Euclidean distance (hclust function of fast- and PLS‑DA cluster package v1.1.25) and six outliers were removed Sparse partial least square-discriminant analyses (sPLS- (SPH_034, SPH_058, SPH_084, SPH_193, SPH_195, DA) [53] were performed with the mixOmics R package and SPH_417). Tus, 76 samples were used for analysis. (v6.10.8) to identify genes that were expressed in blood Block-wise network construction and module detection before vaccination and that were the most discrimina- were carried out with the WGCNA package (v1.68) in tive features of the response to vaccination. Read counts two blocks using the blockwiseModules function with a were fltered, normalized, and corrected for sex, age at threshold power of 6, a height of 0.25, a deep split level weaning, and batch efects, as described for the WGCNA of 2, a reassign threshold of 0.2, and a minimum module analysis. Te classifcation performance (error rate) was size of 30. Te eigenmodules (essentially the frst princi- estimated with the function tune.splsda of the mixOm- pal component of the modules, which can be used as a ics R package. Te tuning was frst performed one com- synthesis often referred to as a “signature” of the module ponent at a time, with a maximum of ncomp = 3 and gene expression) were then correlated with Ab responses. with 5 to 95 (step of 5) genes to test per component, and Modules that were correlated (Pearson correlation) with fvefold cross-validation repeated 100 times. In all analy- the Ab responses with a p-value lower than 0.05 were ses, one component was sufcient to provide the lower considered signifcantly correlated. error rates. Te optimal number of genes to select was then refned by performing another tuning in a more restricted range of genes to test, based on the error rate RNA‑Seq‑based diferential expression analyses of blood of the obtained profles and with a maximum of 25 genes genes between pigs that difered in Ab responses (step of 1), and fvefold cross-validation repeated 100 to vaccine times. Te fnal models included one component and the Diferential expression (DE) analyses were conducted determined number of genes to be selected that led to the using the edgeR package (v3.26.6) in R. Read counts best performance for predicting the classifcation of ani- were fltered, normalized, and corrected for sex, age at mals in high vs low Ab responders at 21, 28, 35, and 118 weaning and batch efects as described for the WGCNA dpv and positive vs negative Ab responders at 21 and 118 analysis. Ten, likelihood ratio tests were performed to dpv. Te set of predictive genes was defned by combin- test, for each gene, the diferential expression between ing all these genes, and PLS-DA were fnally performed extreme groups. Te p-values were adjusted using the to evaluate the predictive capacities of this fnal set of 101 false discovery rate (FDR) method [51] and a signifcance genes. level of 0.05 and a suggestive signifcance level of 0.1 were applied. Results Individual variability of specifc antibody response after M. Feature set enrichment analyses (FSEA) hyo vaccination FSEA were performed with the R package tmod (v0.40) All individual data and metadata are in Additional fle 1: [52]. Hypergeometric tests were used to determine the Table S1. Te M. hyo-specifc Ab response was moni- enrichment of each WGCNA module that was signif- tored by measuring seric M. hyo-specifc IgG levels at cantly correlated with Ab responses (foreground set) in four time points, corresponding to three physiological the total RNA-Seq gene set (background set). Te coinci- steps of the humoral antibody response: early response dent extreme ranks in numerical observations (CERNO) after one injection of vaccine at 21 dpv (49 days of age), method was used to analyze the feature set enrichment maximum intensity of the response after the booster vac- using lists of genes obtained from the DE analyses and cination at 28 and 35 dpv (56 and 63 days of age, respec- ranked by the absolute logarithm of fold change (logFC). tively), and persistence of the Ab response until slaughter To interpret the feature set enrichments, we used gene at 118 dpv (146 days of age) (Fig. 1a). At 21 dpv, M. hyo collections consisting of the blood transcriptomic mod- Ab levels were signifcantly higher in vaccinated than in ules (BTM) and signatures annotated by Li et al. [38], control animals but only one animal reached the thresh- which were adapted by replacing human genes with their old that indicated a M. hyo Ab response (Fig. 1b). Tus, corresponding genes in pigs by Matthijs et al. [43], and at this early time point, we ran the assay with more con- FSEA were visualized with tmodPanelPlot. centrated sera (1/4 dilution, Fig. 1c) to better assess the variability among animals. At this dilution, M. hyo Ab levels were also signifcantly higher in vaccinated than in control animals. M. hyo Ab response reached maximum Blanc et al. Genet Sel Evol (2021) 53:24 Page 7 of 18

intensity at 28 dpv for 79% of animals and at 35 dpv for Additional fle 1: Table S3). Ab levels of animals from the other animals. At 118 dpv, M. hyo Ab levels remained B_1602 were lower than those from B_1611 at 28 dpv above the threshold for 80% of the vaccinated pigs, which and lower than those from B_1604 at 118 dpv (p-val- indicated a persisting humoral response for these ani- ues < 0.05). Please note that females exhibited 30% higher mals. Among the other animals, 8.2 and 11.8% had sus- levels of specifc Ab in response to vaccination at 118 dpv pect or negative M. hyo Ab levels, respectively. than males (see Additional fle 1: Table S3) and Fig. 1d. We observed a high individual variability of the M. hyo Ab levels at each time point (Fig. 1b and Table 1). Te Correlation between M. hyo Ab levels at diferent time coefcients of variation (CV) were equal to 46% at early points (21 dpv, serum dilution 1/4) and late time points (118 M. hyo Ab levels at diferent time points were positively dpv), and 20 and 21% at the maximum response intensity correlated with one another (Fig. 2a). Ab levels at 28 and (28 and 35 dpv). At 21 dpv, the mean values of seric Ab 35 dpv were highly correlated with each other (r = 0.84) levels with the 1/40 dilution were very low compared to and were correlated with Ab levels at 118 dpv (r = 0.62 the other values, with a CV that reached 151% (Table 1). and 0.71, respectively). Ab levels at 21 dpv were moder- In the next steps, for that time point, we used only the Ab ately correlated with later time point responses (r rang- levels measured with the 1/4 dilution, which provided a ing from 0.27 to 0.3). Since M. hyo Ab levels at diferent better range of individual observations. time points can be considered as repeated measure- We assessed the efect of known factors that included ments, temporal correlations were evaluated and were sex, age at weaning, batch, and litter on M. hyo Ab lev- found to be positive, ranging from 0.46 to 0.68 (see Addi- els. Batch had a signifcant (p-values < 0.05) efect at tional fle 1: Table S4). In a principal component analysis 28 and 118 dpv, litter at 21 dpv, and sex at 118 dpv (see (PCA) of M. hyo Ab levels at all time points, animals were

Table 1 Descriptive statistics and heritability estimates of M. hyo Ab vaccine response at diferent days post-vaccination (dpv) Type of Ab response dpv Serum dilution S/P values­ a h2 (SE) Mean SD CV (%)

Early (before booster) 21 1/40 0.041 0.06 151 NC 1/4 0.403 0.19 46 0.57 (0.15) Maximum (post booster) 28 1/40 1.349 0.28 21 0.46 (0.11) 35 1/40 1.215 0.25 20 0.52 (0.11) Persistence (before slaughtering) 118 1/40 0.645 0.30 46 0.52 (0.11)

NC not calculated aAb levels are expressed relatively to a positive control (S/P values, see Methods)

Fig. 2 Correlations of M. hyo Ab levels with production traits. a Correlation matrix (Pearson correlation coefcients) of M. hyo Ab levels in sera at 21, 28, 35, and 118 dpv; BW at 40 and 118 dpv; and ADG from 0 to 40 dpv and 40 to 118 dpv. Color intensity and size of the circle are proportional to the correlation coefcients. Non-signifcant (p > 0.05) correlations are crossed out. b, c Principal component analysis (PCA) of M. hyo Ab levels in sera at 21, 28, 35, and 118 dpv for components b 1–2 and c 1–3 Blanc et al. Genet Sel Evol (2021) 53:24 Page 8 of 18

distributed along the frst component mostly by their Ab growing period (40–118 dpv) were also positively corre- responses at 28, 35 and, 118 dpv and along the second lated (Fig. 2a). As shown in Fig. 2a, Ab levels at 35 dpv component by their Ab responses at 21 dpv (Fig. 2b). Te and BW at 118 dpv or ADG during the growing period two frst components explained 66.5 and 20% of the vari- (from 40 to 118 dpv) were slightly but signifcantly nega- ance, respectively. Te third component separated ani- tively correlated (r = -0.18 and-0.22, respectively). How- mals by their Ab responses at 28 or 35 dpv vs 118 dpv ever, no signifcant correlations were found between the (Fig. 2c). Ab levels at other time points and BW using the whole vaccinated population (Fig. 2a). Comparison of vaccinated and non‑vaccinated groups Ten, we compared BW measurements between ani- for body weight mals with high vs low or negative vs positive Ab response We assessed the efect of vaccination on BW after 40 and by including the corresponding vaccine group as a fxed 118 dpv and on ADG during the post-weaning (0–40 efect in the model. Interestingly, this approach revealed dpv) and growing (40–118 dpv) periods. We applied a that high Ab responders at 28 dpv had 6.4 and 7.3% lower mixed linear model with vaccine group (vaccinated or BW at 40 and 118 dpv, respectively, and had 9.7% and non-vaccinated), sex, and batch as fxed efects, age at 8.3% lower ADG from 0 to 40 dpv and from 40 to 118 weaning and BW at 0 or 40 dpv as linear covariates, and dpv, respectively (Table 2, p < 0.05). In addition, animals litter as a random efect. Sex had a signifcant (p-val- with a positive (vs negative) Ab level at 21 dpv had 5.3% ues < 0.005) efect on BW at 118 dpv and on ADG at lower BW at 40 dpv and 8.1% lower ADG from 0 to 40 40–118 dpv (with higher values for males). Batch had a dpv (p = 0.007). Tus, overall, these results show a slight signifcant efect on BW at 40 dpv and on ADG at 0–40 but signifcantly reduced BW and ADG for the higher Ab dpv. BW at 0 dpv had also a signifcant efect on all BW vaccine responders that could be detected as soon as 40 measurements, with a higher BW at 0 dpv and a higher dpv and persisted until slaughter at 118 dpv. BW or ADG at older ages. Vaccination group had no sig- nifcant efect on BW when the two groups of vaccinated Genetics of vaccine Ab response and non-vaccinated pigs were compared (see Additional Estimates of the heritability of M. hyo Ab response lev- fle 1: Table S5). els were within the same range at all time points (0.46 to 0.57), with a tendency to be slightly higher at the earli- Relationships between M. hyo Ab response and body est time point (Table 1 and Additional fle 1: Table S4). weight in the vaccinated population Estimates of heritabilities and phenotypic and genetic To evaluate the relationships between BW and M. hyo Ab correlations between Ab responses to M. hyo vaccina- levels after vaccination, we performed a correlation anal- tion at the diferent time points are summarized in Addi- ysis (Fig. 2a). BW measurements at diferent time points tional fle 1: Table S6. Estimates of genetic correlations of were positively and signifcantly (p-values < 0.05) cor- Ab levels at 21 dpv with Ab levels at the three later time related throughout the animals’ lifetime. Te ADG dur- points (92 to 98%) were very high. Te lowest correla- ing the post-weaning period (0–40 dpv) and during the tions were detected between Ab levels at 28 and 35, and

Table 2 Estimates of diferences in body weight (BW, kg) at 40 and 118 dpv and in average daily gain (ADG, kg/day) from 0 to 40 dpv and from 40 to 118 dpv between pigs with high vs low or positive vs negative M. hyo Ab response Groups of contrasted Ab responses BW at 40 dpv ADG 0–40 dpv BW at 118 dpv ADG to vaccination at 40–118 dpv dpv Classifcation

21 High vs low 0.102 0.099 0.088 0.194 Positive vs negative 0.007a 0.007a 0.054 0.191 28 High vs low 0.025b 0.025b 0.019b 0.029b 35 High vs low 0.444 0.443 0.113 0.128 118 High vs low 0.259 0.262 0.143 0.182 Positive vs negative 0.833 0.841 0.418 0.310 anegative > positive; blow > high High vs low and positive vs negative animals were defned as described in the Methods section. Details on the assignment of animals to groups at each time point are provided in Additional fle 1: Table S1. BW was ftted using a linear mixed model with sex, batch and groups of contrasted responses to vaccination at each time point (high vs low or positive vs negative) as fxed efects, age at weaning and BW at 0 dpv (weaning) as linear covariates, and litter as random efect. Classifcation p-values of the efect of contrasted groups of vaccine responders at each time point are in this table (signifcant p-values are in italics) Blanc et al. Genet Sel Evol (2021) 53:24 Page 9 of 18

Ab levels at 118 dpv (0.66 and 0.78, respectively). Esti- two diferent time points (21 and 118 dpv), the six other mates of phenotypic correlations were all lower than the modules were correlated at only one time point. All sig- corresponding estimates of genetic correlations, espe- nifcant correlations were low to moderate, with absolute cially for correlations of Ab response at 21 dpv with later values ranging from 0.23 to 0.37. However, each of these time points (see Additional fle 1: Table S6). seven modules was correlated with M. hyo Ab levels in By performing GWAS for each of the Ab pheno- the same orientation at each time point: positive corre- types, we were able to identify two genomic regions lations for purple, sky blue, and brown modules; nega- that were associated with M. hyo Ab levels at 21 dpv tive correlations for dark magenta, dark turquoise, light with sera diluted 1/4 (Fig. 3 and Additional fle 1: cyan, and grey60 modules (Fig. 4a). Te brown module Table S7). Te QTL on Sus scrofa chromosome 1 was positively correlated with vaccine response at 21 (SSC1) SSC1:261,713,894–261,843,495 (129,601 bp dpv (r = 0.25), whereas the dark magenta module was long) included three SNPs with FDR < 0.1 and one negatively correlated (r = − 0.27). Modules that were SNP (AX-116155504) with FDR = 0.047. Te QTL on positively correlated with the M. hyo Ab levels at 28 dpv SSC4, between SSC4:10,201,158 and SSC4:11,076,588 were the purple (r = 0.23) and sky blue (r = 0.31) mod- (875,430 bp long), included 24 SNPs with a FDR < 0.1 and ules. Te three modules that were negatively correlated 51 SNPs with a FDR < 0.05, with AX-116223305 being the with the M. hyo Ab levels at 35 dpv were the dark tur- most strongly associated SNP (FDR = 0.04). Te anno- quoise (r = − 0.27), light cyan (r = − 0.37), and grey60 tated genes that mapped to these QTL are DAB2IP for (r = − 0.24) modules. Only the dark magenta module had the QTL on SSC1 and ASAP1, CYRIB, and GSDMC for a signifcant negative correlation with the M. hyo Ab lev- the QTL on SSC4. els at 118 dpv (r = − 0.27). Five of the seven WGCNA modules that were signif- Correlations of M. hyo Ab levels with co‑expressed genes cantly correlated with Ab levels were enriched in blood in pre‑vaccination blood transcriptomic modules (BTM) defned by Li et al. We performed transcription profling in blood collected [38] (Fig. 4b and Additional fle 1: Table S8). Te dark before vaccination for a subset of 82 pigs, by RNA-Seq magenta module was enriched in the BTM annotated and extracted modules of co-expressed genes using “chromosome Y linked”. Te brown module was enriched WGCNA. We identifed 34 modules that included 31 to in the BTM annotated for heme biosynthesis (M171 and 1606 genes. Seven modules had at least one signifcant M222), cell cycle (M4.1 and M4.2) and dendritic cell/ correlation with M. hyo Ab levels at 21 (two modules), antigen presentation (M87). Te purple module had 28 (two modules), 35 (three modules), and 118 dpv (one functions enriched in BTM annotated for cell migration module) (Fig. 4a, p < 0.05). Only the dark magenta mod- (M45 and M91), natural killer cell enrichment and acti- ule was signifcantly correlated with M. hyo Ab levels at vation (M7.2, MM61.0, M61.2, M157 and signature S1),

Fig. 3 Genome-wide association for M. hyo Ab levels at 21 dpv. M. hyo Ab levels were assessed in sera diluted 1/4. a Manhattan plot based on -log 10 (p-value) from GWAS and imputation analysis against chromosome position annotation of the swine genome assembly 11.1 (Sscrofa11.1). The blue line indicates suggestive association threshold (FDR 0.1) and the red line indicates genome-wide signifcant threshold (FDR 0.05). b QQ = = plot showing the expected distribution of association statistical tests (X-axis) across the SNPs compared to the observed values (Y-axis) Blanc et al. Genet Sel Evol (2021) 53:24 Page 10 of 18

Fig. 4 Weighted gene co-expression network analysis (WGCNA) of modules and M. hyo-specifc Ab response relationships. a Modules extracted by co-expression analyses in blood RNA-Seq dataset (at 0 dpv) are in rows labelled with a color; the number of genes contained in the given module is indicated after the color. Genes that were not assigned to a specifc module (n 689) are grouped in the last module labelled in grey color. The = heatmaps are color-coded according to Pearson’s correlation coefcient between the given module eigengene and the M. hyo Ab levels at 21, 28, 35, and 118 dpv (from positive in blue to negative in orange). Correlation coefcients along with their p-value in parenthesis are mentioned when the p-value < 0.05. b Feature set enrichment analyses (FSEA) of modules correlated with M. hyo-specifc Ab responses. Enrichment was calculated with hypergeometric tests for each WGCNA module (foreground set) among the total RNA-Seq gene set (background set). Only the WGCNA modules that showed a signifcant enrichment (hypergeometric test adjusted p-value < 0.01) in any of the BTM were included. The strength of the p-value is illustrated by a gradient of red. The efect size corresponds to the enrichment calculated as (b/n)/(B/N) where b and B are the numbers of genes from the given module in the fg and bg sets, respectively; n and N, the sizes of the fg and bg sets, respectively. (TBA to be annotated) = and T cell enrichment and activation (M7.0, M7.1, M7.3, dpv between high and low responders (Table 3). A set of M18, M35.0 and M35.1). Te dark turquoise and light 52 genes were diferentially expressed between the two cyan modules had annotated functions in innate antiviral groups of responders at 118 dpv at FDR < 0.1, including a immunity (M13, M68, M75, M111.0, M111.1, M127 and subset of seven genes with FDR < 0.05. M150), infammation/immune response (M24, M37.0, Since the DE analyses revealed only a few DE genes, M78 and M112.0), and activation of dendritic cells functional enrichment analysis was performed using (MS11, M67 and M165). a method that relies on the whole set of analyzed genes that are ranked based on the absolute logFC obtained in Identifcation of diferentially expressed genes in blood the DE analyses. Te functions of genes that had a higher before vaccination between pigs with divergent Ab expression in the pre-vaccination blood transcriptome responses to the M. hyo vaccine of the low responders at 21 dpv were heme biosynthesis Diferential expression analyses compared groups of (M171) and platelet activation (M196 and M199) (Fig. 5 high vs low extreme Ab responders at 21, 28, 35, and and Additional fle 1: Table S9). Genes of the chromo- 118 dpv, and groups of positive vs negative Ab respond- some Y-linked module (M240) had a higher expression ers at 21 and 118 dpv. We identifed four diferentially in the pre-vaccination blood transcriptome of positive expressed genes at 21 dpv between positive and negative responders at 21 dpv. For Ab responses at 28 and 35 Ab responders and one diferentially expressed gene at 28 dpv, genes associated to the same functions had a higher Blanc et al. Genet Sel Evol (2021) 53:24 Page 11 of 18

Table 3 Diferentially expressed genes in blood before vaccination between groups of animals with divergent Ab responses to M. hyo vaccination at diferent days post-vaccination (dpv) Type of Ab response dpv Classifcation Number Diferentially expressed genes with HUGO gene of diferentially nomenclature committee symbol expressed genes Higher expression in "high" Higher expression in "low" or "pos" or "neg"

Early (before booster) 21 High vs low 0 Negative vs positive 4 ­(4a) BARD1a, LMBR1a, TMEM236a, CCDC158a Maximum (post booster) 28 High vs low 1 ­(1a) DNAH9a 35 High vs low 0 Persistence (before 118 High vs low 52 ­(7a) TERB2a, SMPD2, ELOVL1, NCLN, TMTC3a, ICE2a, N4BP2, CCNC, slaughtering) ZNF48, CLCN7, GSDMD, ULK2, LRIF1, AHI1, RPAP3, STK39, ZNF668, TNFRSF18, PTPN18, ABRAXAS1, FAM214A, PIM3, MARVELD1, PHB, ETNK2, TRAM1, ZNF280D, AEBP2, HIST1H4H, ZNF787 ZNF350, TAOK1 Negative vs positive 0 aFDR < 0.05

expression in the pre-vaccination blood transcriptome of were in the DE gene list: NCLN, ENSSSCG00000032640, high responders at both time points. Te functions were CCDC158, TMEM236, DNAH9 and BARD1. In addi- mainly related to innate antiviral immunity (M13, M68, tion, 10 of these predictive genes have been reported to M75, M111.1, M127, and M150) and to dendritic cells be under genetic control in an expression GWAS analy- (M67). For the Ab response at 118 dpv, the genes with a sis performed on 60-day old pig blood transcriptome higher expression in the pre-vaccination blood transcrip- [54]: TXLNB, PPP6R3, ALMS1, PAFAH2, SLC39A7, tome of the positive responders were annotated to mod- CCDC158, MASTL, DNAH9, SENP7, and TBC1D12. ules related to cell junction and adhesion (M1.0, M1.1, and M51), extracellular matrix (M2.0, M2.1 and M2.2), Discussion platelet activation (M85), axon guidance (M110), signal In this study, we characterized the variability of vaccine transduction and infammation (M0 and M82) and mye- humoral response to M. hyo in pigs. We identifed genetic loid cells (M4.3). information and pre-vaccination baseline transcriptomic signatures that could predispose to and predict indi- Blood biomarkers to predict Ab response to M. hyo vaccine vidual M. hyo Ab levels induced after vaccination and We applied the sPLS-DA method [53] to select pre- monitored at diferent time points: early response after dictive genes that can help classify animals as high vs the frst injection (21 dpv), maximum intensity response low Ab responders (at 21, 28, 35, and 118 dpv) or posi- after the booster (28 and 35 dpv), and persistence of Ab tive vs negative Ab responders (at 21 and 118 dpv) (see response until slaughter (118 dpv). Additional fle 1: Table S10). Te detailed contributions Te mechanisms of the vaccine-induced protection to of each gene are in Additional fle 1: Table S11. Interest- control M. hyo infection in pigs are not yet fully under- ingly, the SH3RF1 gene was a predictor of extreme Ab stood. Te M. hyo vaccine is known to induce both local responders at both 35 and 118 dpv. We defned a set of and systemic immune responses that involve specifc Ab 101 candidate predictive genes by grouping the genes production and cellular immunity [5, 55–57]. However, identifed for each time point. PLS-DA were run with the the respective roles of these components of the host 101-gene set to assess its prediction capacity for classi- immune response on the vaccine-induced protection fying animals as high vs low or positive vs negative Ab have not yet been determined and no direct correlation responders at all time points (Fig. 6 and Additional fle 1: has been reported between the variability of any vac- Table S11). A good prediction was obtained with the frst cine response parameter and the protection efciency component (area under the curve (AUC) > 0.86) and a after M. hyo challenge [5, 55, 56]. Vaccine formulations, nearly perfect prediction with two or three components routes of administrations, or adjuvants difered between (AUC > 0.99), with signifcant balanced error rates (BER) these studies and the number of animals studied were for Ab responses at all time points (BER ranging from relatively small (less than 10). Studies that include animal 0.10 to 0.38). Interestingly, among this set of genes, six challenge with the vaccine pathogen after vaccination on Blanc et al. Genet Sel Evol (2021) 53:24 Page 12 of 18

Fig. 5 Feature set enrichment analyses (FSEA) of diferentially expressed (DE) genes between animals with divergent Ab responses to M. hyo vaccination. The CERNO test was operated on lists of genes that were ranked by their absolute log FC of the DE analyses that compare groups of contrasted responders to vaccination at 21, 28, 35, and 118 dpv (high vs low or positive vs negative). BTM signifcantly enriched in each list (p < 0.01 and AUC > 0.75) are represented in rows. BTM are named by their title and ID in parenthesis. The strength of the p-value is illustrated by the transparency of color; the efect size (AUC) is illustrated with the plot size. Signifcantly (p < 0.05) up-regulated genes in high or positive responders are colored in blue and downregulated genes are colored in orange; the others are in grey. (TBA to be annotated) = larger populations are needed to establish the correlates Ab response levels within an animal were positively of vaccination protection. In our study, we focused on correlated with each other across time points, which the humoral response following vaccination and found a revealed the absence of an antagonism between Ab high inter-individual variability in M. hyo Ab levels, infu- response levels at diferent time points. Selecting for enced by both the host genetics and blood transcriptom- Ab response intensity to vaccination may have an ics. Tus, we provide a proof of concept that genetic and impact on pig production performance due to trade- blood transcriptomic data collected before vaccination of issues. Comparisons between vaccinated and non- are relevant resources to predict vaccine Ab response lev- vaccinated animal groups showed no diferences in els, which could be applied to the correlates of protection BW, which highlights that the vaccine has no negative that need to be established in future studies. efect on BW at the whole population level. In feld Blanc et al. Genet Sel Evol (2021) 53:24 Page 13 of 18

Fig. 6 Partial least square-discriminant analysis (PLS-DA) with the 101 predictive genes for M. hyo vaccine response. Projection of the groups of contrasted responders to vaccination at 21, 28, 35, and 118 dpv (high vs low or positive vs negative) into the subspace spanned by the frst two components after a PLS-DA analysis with the 101 candidate predictive genes. Confdence ellipses for each class are plotted (confdence level set to 95%)

conditions where M. hyo is circulating, vaccinated pigs throughout life, rather than selecting for maximal Ab have higher growth rates [5, 6], due to the herd protec- response capacity. Correlates of protection are still tion against the pathogen. In our study, in which no lacking to be able to determine what threshold of Ab M. hyo infection occurred, we observed negative cor- level would be sufcient to optimize protection while relations between Ab responses to M. hyo vaccination limiting the decrease in BW. When vaccine response and BW. Te high vaccine responders had signifcantly traits are included in breeding programs, a joint assess- lower BW and ADG values than the low Ab respond- ment of functional trade-ofs and economic impacts ers, which suggests a trade-of. Tus, breeding goals will be necessary to optimize both vaccine efectiveness should aim at reaching levels of Ab that are sufcient and production traits. Te balance between sustainabil- for an efcient protection of animals that persists ity, feasibility and desirability of breeding livestock for Blanc et al. Genet Sel Evol (2021) 53:24 Page 14 of 18

disease resistance remains a main issue that needs to be booster vaccination [42]. It should be noted that, in the addressed [58]. same population, the kinome (global cellular kinase activ- From the blood transcriptome profles obtained prior ity) revealed prior vaccination candidate biomarkers [32]. vaccination, we identifed diferentially expressed genes In our study, we investigated the early, maximum inten- in animals that showed divergent levels of Ab response to sity, and persistence of Ab responses and searched for vaccination, especially for persistence of the Ab response. predictors of the intensity of vaccine response by study- Although these genes were not shared with previously ing genes that are expressed in pre-vaccination blood. reported gene signatures associated with early response Merging the best predictors of Ab responses at each to vaccination in humans or pigs [35, 37, 42], they were time point led to a set of 101 genes (see Additional fle 1: enriched in blood cell functions that are biologically rele- Table S11) that had an enhanced performance of predic- vant to inter-individual variability of Ab response to vac- tion at all time points compared to single subsets of genes cination. Te infammation, innate antiviral immunity, defned at each time point. A sPLS-DA analysis does not platelet activation, dendritic cells/antigen presentation, necessarily select genes that are of biological relevance myeloid cells, natural killer cells, and T cell activation to the predicted phenotype. However, six of the 57 genes functions associated with the Ab response that we have that were diferentially expressed between the high and identifed in our study have already been reported as low responders were found in this set of 101 genes that gene signatures of vaccination in humans [38, 59], pigs were predictive of Ab responses. Tese six genes (NCLN, [43], and sheep [39, 40]. ENSSSCG00000032640, CCDC158, TMEM236, DNAH9, Te pre-vaccination blood transcriptome predicted and BARD1) were among the best contributors to the M. hyo Ab levels at the early and maximum intensity prediction. More studies are needed to refne the list of responses, and strikingly at a late time point that cor- candidate predictors and to validate them in other popu- responded to Ab persistence. Interestingly, the blood lations, with diferent genetic setups and environments of cell functions involved in predisposition to vaccination production. Since more and more studies aim at identify- response varied between time points, which suggests ing genetic predictors and, more generally, biomarkers of that the underlying biological mechanisms involved in vaccine efectiveness in pigs, it will be essential to analyze the prediction probably difer, and their study needs to whether such predictors apply to diferent vaccines. Inte- be deepened. A high early response was associated with a grating various layers of biological information from the lower expression in blood of genes that are related to cell host (genetics, transcripomics, kinomics, etc.) and its gut cycle and transcription, heme biosynthesis, and platelet microbiota is expected to create a breakthrough in our activation functions. High responders at the maximum understanding of the determinants of vaccine efective- intensity of Ab response showed a higher expression of ness and more globally on immune capacity and animal genes related to dendritic cell, natural killer and T cell robustness, with potential applications to improve the activation and antiviral and innate immune response sustainability of pig production systems. (mainly interferon signature and RIG-1 like sensing), as In this study, an inter-individual variability of Ab lev- well as complement activation. Te persistence of the Ab els in response to M. hyo vaccination was reported, with response was associated with genes related to cell adhe- females exhibiting higher levels of Ab than males (uncas- sion, extracellular matrix, platelet activation and mono- trated) at the latest time point. Tis is consistent with the cyte signature. Finally, high Ab responses at nearly all correlation that we found between a male-specifc mod- time points were associated with a decreased represen- ule (M240) and Ab response that was previously reported tation of genes associated to the chromosome Y-specifc in humans for other vaccines [38], and highlights a gen- module (M240) compared to low Ab responses. All these der bias in Ab production. Sex diferences have been blood cell functions have physiological roles that support described in immunity to multiple vaccines, including their implication in the immune capacity of animals to both inactivated and live vaccines in humans, with Ab mount an Ab response to vaccination. Te pig Y chromo- responses often being higher in females than in males some harbors genes with regulatory properties that may [64, 65]. Tus, sex diferences in immune responses to shape the immune cell transcriptomes [60]. Heme regu- vaccines that could be caused by genetic, hormonal, lates B-cell diferentiation and antibody class switch [61, microbiota, and environmental factors, or a combina- 62] and, in general, an altered CD4 T cell signature can tion of these, should be taken into account in vaccination be used as a predictive immune phenotype for low vac- studies. cine responsiveness in middle-aged humans [63]. In humans, estimates of heritability of Ab responses A previous study in pigs did not identify any biomark- to several vaccinations ranged from low (< 0.20) to high ers in the blood transcriptome before vaccination for the (> 0.80), depending on the vaccine [23, 66–69]. In pigs, prediction of M. hyo Ab levels measured 10 days after a mainly low to moderate heritabilities of specifc Ab Blanc et al. Genet Sel Evol (2021) 53:24 Page 15 of 18

production have been reported [29, 30]. In a previous were predictive of Ab responses in our study are under study on Large White pigs, we showed that M. hyo-spe- genetic control [54]. Deciphering the genetic control of cifc Ab response after a one-shot injection of vaccine the expression of genes in blood should also help iden- had a large phenotypic variance, with a low heritability of tify the best variant predictors of vaccine responses that 0.12 [31]. Here, we found heritabilities of M. hyo-specifc could be implemented in breeding programs. Of note, we Ab levels ranging from 0.46 to 0.57, with the higher value also recently reported an impact of the fecal microbiota for the early M. hyo Ab levels at 21 dpv. Tese heritability collected before vaccination on vaccine efectiveness in values might be slightly overestimated, as common envi- the same population [79], in agreement with fndings on ronmental efect of birth litter was not included in the another pig population [42]. Te accumulation of data on fnal model of analysis due to numerical issues. However, genetic variability and predictors to vaccine responses the variance due to common litter efects rarely exceeds (transcriptome, kinome, microbiome) should provide 10% of the phenotypic variance, so that heritability values new insights to improve vaccination efcacy. From a of 0.46 to 0.57 are very likely to be mainly due to additive One-Health perspective, precision livestock production genetic variation. GWAS revealed two signifcant peaks and personalized medicine share the same requirements for early M. hyo Ab levels at 21 dpv, but no associations at to better understand the determinants of individual base- other time points. Tus, we confrm the infuence of the line variability that shape immune competence, and to host genetics on vaccine efectiveness but larger popula- assess if such knowledge will be useful in future breeding tion sizes are needed to go deeper in the analysis of the practices. underlying genetic determinisms, which likely involves numerous genetic loci spread along the genome. In addi- Conclusions tion, we anticipate that biological relays that are con- We showed and thus confrmed a genetic basis for the trolled by diferent genetic determinisms are put in place variability of Ab vaccine response in pigs vaccinated during the time course of the vaccine response. against M. Hyo. Heritability of the Ab response at all time Te few genes that map to the two GWAS peaks have points could be estimated and genomic regions asso- functional links with immunity in humans. On SSC1, the ciated with the early response (21 dpv) identifed. We DAB2IP gene encodes the DAB2 interacting protein that also identifed individual baseline blood transcriptomic functions as a scafold protein involved in numerous pro- signatures prior to vaccination that predicted high or cesses including innate immune response, infammation low responses in the tested population. Similar genetic and cell cycle. A GWAS in humans revealed associations and genomic studies could be applied to the correlates of this gene with aggressive prostate cancer, DAB2IP of protection that still need to be established. Tese being a candidate tumor suppressor gene [70]. DAB2IP results highlight that the variability of baseline genomic is also known to inhibit tumor cell growth in vitro and expression beyond SNPs may be relevant to predict the in vivo, which highlights important properties for anti- variability of vaccine efectiveness and baseline immune cancer therapies [71]. On SSC4, the ASAP1 gene (Arf- competence. GAP with SH3 domain, ankyrin repeat and PH domain 1) encodes an ADP-ribosylation factor GTPase-activating Supplementary Information protein. ASAP1 plays a role in regulating the migration The online version contains supplementary material available at https​://doi. org/10.1186/s1271​1-021-00614​-5. of dendritic cells, with a possible impact on predisposi- tion to tuberculosis [72]. Associations of this gene with Additional fle 1: Table S1. Individual data and metadata. Table S2. mean platelet volume and platelet counts have also been Descriptive statistics of the groups of animals with contrasted Ab reported [73, 74]. Te CYRIB gene (CYFIP related Rac1 response to vaccination against M. hyo at diferent days post-vaccination interactor B) is related with several pathways, among (dpv) within the total population and the subset for which RNASeq data are available. Table S3. Efects of sex, batch, age at weaning and litter which response to an elevated concentration of platelet on M. hyo Ab levels. Table S4. Correlations of Ab responses to M. hyo cytosolic Ca2+. GWAS in humans have revealed asso- vaccination at diferent days post-vaccination (dpv). Table S5. Efects of ciations of variants of regulatory elements of CYRIB with sex, batch, age at weaning, BW at 0 dpv, vaccination and litter on BW at 40 and 118 dpv and ADG from 0 to 40 and 40 to 118 dpv. Table S6. Genetic lymphocyte counts [75]. Te GSDMC gene (gasdermin parameters of Ab responses to M. hyo vaccination at diferent days C) was initially identifed as a gene that is preferentially post-vaccination (dpv). Table S7. Suggestive (FDR < 0.1) and signifcant expressed in metastatic melanoma cells [76] and also in (FDR < 0.05) SNPs associated with the early M. hyo Ab levels at 21 dpv (sera diluted 1/4). Table S8. FSEA of transcript overlap between WGCNA mod‑ the epithelium of the skin and gastrointestinal tract [77], ules and BTM. Table S9. FSEA of diferentially expressed genes between while GWAS in humans revealed associations of variants groups of animals with contrasted Ab responses to M. hyo vaccination. in GSDMC with monocyte counts [78]. Table S10. Determination of the number of genes (maximum 25) neces‑ sary to reach an optimized prediction of extreme Ab responders to M. hyo Finally, an eGWAS performed on 60-day old pig blood vaccination by sPLS-DA. Table S11. Contribution to the sPLS-DA analysis transcriptome revealed that 10 of the 101 genes that Blanc et al. Genet Sel Evol (2021) 53:24 Page 16 of 18

References of the diferent phenotypes of the set of 101 genes predictive of extreme 1. Posteraro B, Pastorino R, Di Giannantonio P, Ianuale C, Amore R, Ric‑ Ab responders to M. hyo vaccination. Table S12. PLS-DA with the set of ciardi W, et al. The link between genetic variation and variability in 101 genes predictive of extreme Ab responders to M. hyo vaccination. vaccine responses: systematic review and meta-analyses. Vaccine. 2014;32:1661–9. 2. Jorge S, Dellagostin OA. The development of veterinary vaccines: a Acknowledgements review of traditional methods and modern biotechnology approaches. The authors would like to thank Isabelle Schwartz-Cornil who coordinated the Biotechnol Res Innov. 2017;1:6–13. H2020 SAPHIR project that gathered a large multidisciplinary consortium and 3. Maes D, Sibila M, Kuhnert P, Segalés J, Haesebrouck F, Pieters M. Update gave us an opportunity to design ad hoc experiments to study individual vari‑ on Mycoplasma hyopneumoniae infections in pigs: knowledge gaps for ability of vaccine responses. We have special warm thanks to Chiara Ferrandi improved disease control. Transbound Emerg Dis. 2018;65:110–24. (†) and Alessandra Stella who helped producing the libraries for RNA-Seq at 4. Holst S, Yeske P, Pieters M. Elimination of Mycoplasma hyopneumoniae Parco Technologico (PTP, Lodi, Italy). We also thank Rémy-Félix Serre (INRAE, from breed-to-wean farms: a review of current protocols with emphasis Get-Plage platform) for his contribution to the RNA-Seq library sequencing. on herd closure and medication. J Swine Health Prod. 2015;23:321–30. We thank the INRAE Gentyane Genomic Services platform (Clermont-Ferrand, 5. Maes D, Deluyker H, Verdonck M, Castryck F, Miry C, Vrijens B, et al. Efect France) who performed the animal SNP genotyping. We are grateful to of vaccination against Mycoplasma hyopneumoniae in pig herds with an Hélène Hayes (INRAE GABI unit) for a much appreciated proofreading of the all-in/all-out production system. Vaccine. 1999;17:1024–34. manuscript. We thank all members of the Genetics, Microbiota, Health team of 6. Siugzdaite J, Garlaite K, Urbsiene D. Evaluation of antibody formation, the INRAE GABI unit for their help at all steps of the project, including animal daily weight gain and meat quality after vaccination of piglets. Acta Vet sampling. We thank Andrea Rau and Florence Jafrezic (INRAE GABI unit) for Hung. 2003;51:273–81. their helpful advice on statistical analyses. We also thank the pig team of the 7. Meyns T, Dewulf J, de Kruif A, Calus D, Haesebrouck F, Maes D. Compari‑ GenESI unit for animal housing and sampling all along the experiment. son of transmission of Mycoplasma hyopneumoniae in vaccinated and non-vaccinated populations. Vaccine. 2006;24:7081–6. Authors’ contributions 8. Sibila M, Nofrarias N, Lopez-Soria S, Segalés J, Valero O, Espinal A, et al. FB, JE, MHPvdL and CRG conceived and designed the experimental design; YB Chronological study of Mycoplasma hyopneumoniae infection, serocon‑ and LR were in charge of the pig production in the experimental farm; JJL and version and associated lung lesions in vaccinated and non-vaccinated GL managed animal sampling; GL organized the biobanking of all samples; OB pigs. Vet Microbiol. 2007;122:97–107. produced the RNA sequencing data; FB characterized the vaccine response; 9. Villarreal I, Meyns T, Dewulf J, Vranckx K, Calus D, Pasmans F, et al. The JE and TM performed bioinformatics analyses with the help of BBH; FB, JE and efect of vaccination on the transmission of Mycoplasma hyopneumoniae MR performed GWAS analyses with the help of BBH; JPB performed heritability in pigs under feld conditions. Vet J. 2011;188:48–52. analyses; FB and JE performed bio-statistical analyses; CRG supervised the 10. Poland GA, Kennedy RB, Mckinney BA, Ovsyannikova IG, Lambert ND, project; FB and CRG interpreted all data and wrote the paper. All authors read Jacobson RM, et al. Vaccinomics, adversomics, and the immune response and approved the fnal manuscript. network theory: Individualized vaccinology in the 21st century. Semin Immunol. 2013;25:89–103. Funding 11. Raeven RHM, van Riet E, Meiring HD, Metz B, Kersten GFA. Systems vac‑ This project has received funding from the European Union’s Horizon 2020 cinology and big data in the vaccine development chain. Immunology. Programme for research, technological development and demonstration 2018;156:33–46. under the Grant Agreement n°633184. This publication refects the views only 12. Lambert ND, Haralambieva IH, Kennedy RB, Ovsyannikova IG, Pankratz VS, of the authors, and not the European Commission (EC). The EC is not liable for Poland GA. Polymorphisms in HLA-DPB1 are associated with diferences any use that may be made of the information contained herein. in Rubella virus-specifc humoral immunity after vaccination. J Infect Dis. 2015;211:898–905. Availability of data and materials 13. Kennedy RB, Ovsyannikova IG, Haralambieva IH, Lambert ND, Pankratz The blood transcriptome RNA-Seq datasets generated and analysed during VS, Poland GA. Genome-wide SNP associations with Rubella-specifc the current study are available in the SRA repository with accession number cytokine responses in measles-mumps-rubella vaccine recipients. Immu‑ PRJNA661433. All other data generated or analysed during this study are nogenetics. 2014;66:493–9. included in this published article and its additional fles. 14. Voigt EA, Haralambieva IH, Larrabee BL, Kennedy RB, Ovsyannikova IG, Schaid DJ, et al. Polymorphisms in the Wilms tumor gene are associated Ethics approval and consent to participate with interindividual variations in Rubella virus-specifc cellular immunity All animal procedures were performed according to the guidelines for the after measles-mumps-rubella II vaccination. J Infect Dis. 2018;217:560–6. care and use of experimental animals established by INRAE (ability for animal 15. Haralambieva IH, Kennedy RB, Ovsyannikova IG, Whitaker JA, Poland GA. experimentation to C. Rogel-Gaillard: A78-172; agreement for experimenta‑ Variability in humoral immunity to measles vaccine: New developments. tion of Centre de recherche INRA Poitou–Charentes—Site expérimental du Trends Mol Med. 2015;21:789–801. Magneraud: A17661; protocol approved by the French Ministry of Research 16. Scepanovic P, Alanio C, Hammer C, Hodel F, Bergstedt J, Patin E, et al. with authorization ID APAFIS#4295–2016022615583351 v4 after the review of Human genetic variants and age are the strongest predictors of humoral ethics committee Nº084). immune responses to common pathogens and vaccines. Genome Med. 2018;10:59. Consent for publication 17. Wu T, Chen C, Lai S, Lin HH, Chu C, Wang L. SNP rs7770370 in HLA-DPB1 Not applicable. loci as a major genetic determinant of response to booster hepatitis B vaccination: results of a genome-wide association study. J Gastroenterol Competing interests Hepatol. 2015;30:891–9. The authors declare that they have no competing interests. 18. Chung S, Roh EY, Park B, Lee Y, Shin S, Yoon JH, et al. GWAS identifying HLA-DPB1 gene variants associated with responsiveness to hepatitis B Author details virus vaccination in Koreans: Independent association of HLA-DPB1*04:02 1 Université Paris Saclay, INRAE, AgroParisTech, GABI, 78350 Jouy en Josas, ‐ ‐ ‐ possessing rs1042169 G–rs9277355 C–rs9277356 A. J Viral Hepat. France. 2 INRAE, GenESI, 17700 Surgères, France. 3 INRAE, GeT-PlaGe, Genotoul, 2019;26:1318–29. 31326 Castanet‑Tolosan, France. 19. Pan L, Zhang L, Zhang W, Wu X, Li Y, Yan B, et al. A genome-wide associa‑ tion study identifes polymorphisms in the HLA-DR region associated Received: 6 August 2020 Accepted: 16 February 2021 with non-response to hepatitis B vaccination in Chinese Han populations. Hum Mol Genet. 2014;23:2210–9. 20. Linnik JE, Egli A. Impact of host genetic polymorphisms on vaccine induced antibody response. Hum Vaccin Immunother. 2016;12:907–15. Blanc et al. Genet Sel Evol (2021) 53:24 Page 17 of 18

21. Ovsyannikova IG, Kennedy RB, O’Byrne M, Jacobson RM, Pankratz VS, microbiota in high and low responder pigs before, and after vaccination Poland GA. Genome-wide association study of antibody response to for Mycoplasma hyopneumoniae. Vaccine. 2019;37:1743–55. smallpox vaccine. Vaccine. 2012;30:4182–9. 43. Matthijs AMF, Auray G, Jakob V, García-Nicolás O, Braun RO, Keller I, et al. 22. Pajewski NM, Shrestha S, Quinn CP, Parker SD, Wiener H, Aissani B, et al. A Systems immunology characterization of novel vaccine formulations for genome-wide association study of host genetic determinants of the anti‑ Mycoplasma hyopneumoniae bacterins. Front Immunol. 2019;10:1087. body response to anthrax vaccine adsorbed. Vaccine. 2012;30:4778–84. 44. R Core Team. R: a language and environment for statistical computing. 23. O’Connor D, Png E, Khor CC, Snape MD, Hill AVS, van der Klis F, et al. Vienna: R Foundation for Statistical Computing; 2019. https​://www.r-proje​ Common genetic variations associated with the persistence of immunity ct.org/ following childhood immunization. Cell Rep. 2019;27:e4. 45. Bates D, Mächler M, Bolker B, Walker S. Fitting linear mixed-efects models 24. Leach RJ, Neill RGO, Fitzpatrick JL, Williams JL, Glass EJ. Quantitative trait using lme4. J Stat Softw. 2015;67:1–48. loci associated with the immune response to a bovine respiratory syncyt‑ 46. Kuznetsova A, Brockhof PB, Christensen RHB. lmerTest package: tests in ial virus vaccine. PLoS One. 2012;7:e33526. linear mixed efects models. J Stat Softw. 2017;82:1–26. 25. Luo C, Qu H, Ma J, Wang J, Li C, Yang C, et al. Genome-wide association 47. Lê S, Josse J, Husson F. FactoMineR: a package for multivariate analysis. J study of antibody response to Newcastle disease virus in chicken. BMC Stat Softw. 2008;25:1–18. Genomics. 2013;14:42. 48. Groeneveld E, Kovac M, Mielenz N. VCE User’s guide and reference 26. Rothschild M, Hill H, Christian L, Warner C. Genetic diferences in serum- manual. Version 6.0. Neustadt: Institute of Farm Animal Genetics; 2008. neutralization titers of pigs after vaccination with pseudorabies modifed 49. Rönnegård L, McFarlane SE, Husby A, Kawakami T, Ellegren H, Qvarnström live-virus vaccine. Am J Vet Res. 1984;45:1216–8. A. Increasing the power of genome wide association studies in natural 27. Zanella R, Gava D, de Oliveira J, Schaefer R, Ciacci-Zanella JR, Biondo populations using repeated measures—evaluation and implementation. N, et al. Unravelling the genetic components involved in the immune Methods Ecol Evol. 2016;7:792–9. response of pigs vaccinated against infuenza virus. Virus Res. 50. Langfelder P, Horvath S. WGCNA: an R package for weighted correlation 2015;210:327–36. network analysis. BMC Bioinformatics. 2008;9:559. 28. Dunkelberger JR, Serão NVL, Weng Z, Waide EH, Niederwerder MC, Kerri‑ 51. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practi‑ gan MA, et al. Genomic regions associated with host response to porcine cal and powerful approach to multiple testing. J R Stat Soc Ser B. reproductive and respiratory syndrome vaccination and co-infection in 1995;57:289–300. nursery pigs. BMC Genomics. 2017;18:865. 52. Weiner J 3rd, Domaszewska T. tmod : an R package for general and multi‑ 29. Uddin MJ, Grosse-Brinkhaus C, Cinar MU, Jonas E, Tesfaye D, Tholen E, variate enrichment analysis. PeerJ Prepr. 2016;4:e2420v1. et al. Mapping of quantitative trait loci for mycoplasma and tetanus anti‑ 53. Lê Cao K-A, Boitard S, Besse P. Sparse PLS discriminant analysis : biologi‑ bodies and interferon-gamma in a porcine F­ 2 Duroc Pietrain resource cally relevant feature selection and graphical displays for multiclass population. Mamm Genome. 2010;21:409–18. × problems Sparse PLS discriminant analysis: biologically relevant feature 30. Edfors-Lilja I, Wattrang E, Marklund L, Moller M, Andersson-Eklund L, selection and graphical displays for multiclass problems. BMC Bioinfor‑ Andersson L, et al. Mapping quantitative trait loci for immune capacity in matics. 2011;12:253. the pig. J Immunol. 1998;161:829–35. 54. Maroilley T, Lemonnier G, Lecardonnel J, Esquerré D, Ramayo-Caldas Y, 31. Flori L, Gao Y, Laloë D, Lemonnier G, Leplat J, Teillaud A, et al. Immunity Mercat MJ, et al. Deciphering the genetic regulation of peripheral blood traits in pigs: substantial genetic variation and limited covariation. PLoS transcriptome in pigs through expression genome-wide association One. 2011;6:e22717. study and allele-specifc expression analysis. BMC Genomics. 2017;18:967. 32. Lipsit SWL, Wilkinson J, Scruten E, Facciuolo A, Denomy C, Griebel PJ, et al. 55. Djordjevic S, Eamens GJ, Romalis LF, Nicholls PJ, Taylor V, Chin J. Serum Kinome profling of peripheral blood mononuclear cells collected prior and mucosal antibody responses and protection in pigs vaccinated to vaccination reveals biomarkers and potential mechanisms of vaccine against Mycoplasma hyopneumoniae with vaccines containing a dena‑ unresponsiveness in pigs. Sci Rep. 2020;10:11546. tured membrane antigen pool and adjuvant. Aust Vet J. 1997;75:504–11. 33. Chaussabel D, Pascual V, Banchereau J. Assessing the human immune 56. Thacker EL, Thacker BJ, Boettcher TB, Jayappa H. Comparison of antibody system through blood transcriptomics. BMC Biol. 2010;8:84. production, lymphocyte stimulation, and protection induced by four 34. Mach N, Gao Y, Lemonnier G, Lecardonnel J, Oswald IP, Estellé J, et al. commercial Mycoplasma hyopneumoniae bacterins. Swine Health Prod. The peripheral blood transcriptome refects variations in immunity 1998;6:107–12. traits in swine: Towards the identifcation of biomarkers. BMC Genomics. 57. Marchioro SB, Maes D, Flahou B, Pasmans F, Del Pozo SR, Vranckx K, et al. 2013;14:894. Local and systemic immune responses in pigs intramuscularly injected 35. Gonçalves E, Bonduelle O, Soria A, Loulergue P, Rousseau A, Cachanado with an inactivated Mycoplasma hyopneumoniae vaccine. Vaccine. M, et al. Innate gene signature distinguishes humoral versus cytotoxic 2013;31:1305–11. responses to infuenza vaccination. J Clin Invest. 2019;129:1960–71. 58. Stear MJ, Bishop SC, Mallard BA, Raadsma H. The sustainability, feasibility 36. Querec TD, Akondy RS, Lee EK, Cao W, Nakaya HI, Teuwen D, et al. Systems and desirability of breeding livestock for disease resistance. Res Vet Sci. biology approach predicts immunogenicity of the yellow fever vaccine in 2001;71:1–7. humans. Nat Immunol. 2009;10:116–25. 59. Tsang JS, Schwartzberg PL, Kotliarov Y, Biancotto A, Xie Z, Germain RN, 37. Ovsyannikova IG, Salk HM, Kennedy RB, Haralambieva IH, Zimmermann et al. Global analyses of human immune variation reveal baseline predic‑ MT, Grill DE, et al. Gene signatures associated with adaptive humoral tors of postvaccination responses. Cell. 2014;157:499–513. immunity following seasonal infuenza A/H1N1 vaccination. Genes 60. Case LK, Wall EH, Dragon JA, Saligrama N, Krementsov DN, Moussawi M, Immun. 2016;17:371–9. et al. The Y chromosome as a regulatory element shaping immune cell 38. Li S, Rouphael N, Duraisingham S, Romero-Steiner S, Presnell S, Davis C, transcriptomes and susceptibility to autoimmune disease. Genome Res. et al. Molecular signatures of antibody responses derived from a systems 2013;23:1474–85. biology study of fve human vaccines. Nat Immunol. 2014;15:195–204. 61. Tsui C, Martinez-Martin N, Gaya M, Maldonado P, Llorian M, Legrave NM, 39. Jouneau L, Lefebvre DJ, Costa F, Romey A, Blaise-Boisseau S, Relmy A, et al. Protein kinase C-β dictates B cell fate by regulating mitochondrial et al. The antibody response induced FMDV vaccines in sheep correlates remodeling, metabolic reprogramming, and heme biosynthesis. Immu‑ with early transcriptomic responses in blood. NPJ Vaccines. 2020;5:1. nity. 2018;48:e5. 40. Braun RO, Brunner L, Wyler K, Auray G, García-Nicolás O, Python S, et al. 62. Watanabe-Matsui M, Muto A, Matsui T, Itoh-Nakadai A, Nakajima O, System immunology-based identifcation of blood transcriptional mod‑ Murayama K, et al. Heme regulates B-cell diferentiation, antibody class ules correlating to antibody responses in sheep. NPJ Vaccines. 2018;3:41. switch, and heme oxygenase-1 expression in B cells as a ligand of Bach2. 41. Adler M, Murani E, Ponsuksili S, Wimmers K. PBMC transcription profles Blood. 2011;117:5438–48. of pigs with divergent humoral immune responses and lean growth 63. van der Heiden M, Berbers GAM, Fuentes S, van Zelm MC, Boots AMH, performance. Int J Biol Sci. 2013;9:907–16. Buisman A. An explorative biomarker study for vaccine responsiveness 42. Munyaka PM, Kommadath A, Fouhse J, Wilkinson J, Diether N, Stothard P, after a primary meningococcal vaccination in middle-aged adults. Front et al. Characterization of whole blood transcriptome and early-life fecal Immunol. 2018;8:1962. Blanc et al. Genet Sel Evol (2021) 53:24 Page 18 of 18

64. Klein SL, Flanagan KL. Sex diferences in immune responses. Nat Rev 74. Meisinger C, Prokisch H, Gieger C, Soranzo N, Mehta D, Rosskopf D, et al. Immunol. 2016;16:626–38. A genome-wide association study identifes three loci associated with 65. Klein SL, Marriott I, Fish EN. Sex-based diferences in immune function mean platelet volume. Am J Hum Genet. 2009;84:66–71. and responses to vaccination. Trans R Soc Trop Med Hyg. 2014;109:9–15. 75. Astle WJ, Elding H, Jiang T, Allen D, Ruklisa D, Mann AL, et al. The allelic 66. Newport MJ, Goetghebuer T, Weiss HA, Whittle H, Siegrist C-A, Marchant landscape of human blood cell trait variation and links to common A. Genetic regulation of immune responses to vaccines in early life. complex disease. Cell. 2016;167:e19. Genes Immun. 2004;5:122–9. 76. Watabe K, Ito A, Asada H, Endo Y, Kobayashi T, Nakamoto K, et al. Struc‑ 67. Tan P, Jacobson RM, Poland GA, Jacobsen SJ, Pankratz VS. Twin studies ture, expression and chromosome mapping of MLZE, a novel gene which of immunogenicity—determining the genetic contribution to vaccine is preferentially expressed in metastatic melanoma cells. Jpn J Cancer Res. failure. Vaccine. 2001;19:2434–9. 2001;92:140–51. 68. Höhler T, Reuss E, Evers N, Dietrich E, Rittner C, Freitag CM, et al. Dif‑ 77. Tamura M, Tanaka S, Fujii T, Aoki A, Komiyama H, Ezawa K, et al. Members ferential genetic determination of immune responsiveness to hepatitis of a novel gene family, Gsdm, are expressed exclusively in the epithelium B surface antigen and to hepatitis A virus: a vaccination study in twins. of the skin and gastrointestinal tract in a highly tissue-specifc manner. Lancet. 2002;360:991–5. Genomics. 2007;89:618–29. 69. Schaid DJ, Haralambieva IH, Larrabee BR, Ovsyannikova IG, Kennedy RB, 78. Crosslin DR, McDavid A, Weston N, Zheng X, Hart E, de Andrade M, et al. Poland GA, et al. Heritability of vaccine-induced measles neutralizing Genetic variation associated with circulating monocyte count in the antibody titers. Vaccine. 2017;35:1390–4. eMERGE network. Hum Mol Genet. 2013;22:2119–27. 70. Duggan D, Zheng SL, Knowlton M, Benitez D, Dimitrov L, Wiklund F, et al. 79. Munyaka PM, Blanc F, Estellé J, Lemonnier G, Leplat J-J, Rossignol M-N, Two genome-wide association studies of aggressive prostate cancer et al. Discovery of predictors of Mycoplasma hyopneumoniae vaccine implicate putative prostate tumor suppressor gene DAB2IP. J Natl Cancer response efciency in pigs: 16S rRNA gene fecal microbiota analysis. Inst. 2007;99:1836–44. Microorganisms. 2020;8:1151. 71. Liu L, Xu C, Hsieh JT, Gong J, Xie D. DAB2IP in cancer. Oncotarget. 2016;7:3766–76. 72. Curtis J, Luo Y, Zenner HL, Cuchet-Lourenço D, Wu C, Lo K, et al. Suscepti‑ Publisher’s Note bility to tuberculosis is associated with variants in the ASAP1 gene encod‑ Springer Nature remains neutral with regard to jurisdictional claims in pub‑ ing a regulator of dendritic cell migration. Nat Genet. 2015;47:523–7. lished maps and institutional afliations. 73. Qayyum R, Snively BM, Ziv E, Nalls MA, Liu Y, Tang W, et al. A meta-analysis and genome-wide association study of platelet count and mean platelet volume in African Americans. PLoS Genet. 2012;8:e1002491.

Ready to submit your research ? Choose BMC and benefit from:

• fast, convenient online submission • thorough peer review by experienced researchers in your field • rapid publication on acceptance • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations • maximum visibility for your research: over 100M website views per year

At BMC, research is always in progress.

Learn more biomedcentral.com/submissions